Can you please resubmit this with git format-patch, if still
applicable on FreeBSD?
Avati
On Tue, Feb 17, 2009 at 1:49 AM, Filipe Maia <filipe@xxxxxxxxxxxxxx> wrote:
> Hi,
>
> The previous patch does not compile in FreeBSD as I used a non
> standard part of the stat structure.
>
> Here's the fix against 2.0.0r18 to make it compile in FreeBSD (and
> possibly other POSIX systems).
>
> --- posix.c.orig 2009-02-16 21:10:36.963597828 +0100
> +++ posix.c 2009-02-16 21:11:05.587603273 +0100
> @@ -2864,8 +2864,8 @@
> */
>
> if (!S_ISLNK (entry->buf.st_mode)) {
> - tv[0].tv_sec = entry->buf.st_atim.tv_sec;
> - tv[1].tv_sec = entry->buf.st_mtim.tv_sec;
> + tv[0].tv_sec = entry->buf.st_atime;
> + tv[1].tv_sec = entry->buf.st_mtime;
> ret = utimes (pathname, tv);
> if (ret == -1) {
> op_ret = -errno;
>
> Filipe
>
> On Sat, Jan 31, 2009 at 06:09, Anand Avati <avati@xxxxxxxxxxxxx> wrote:
>> Committed (with some minor changes). Thanks! There is no sensible way
>> to set ctime.
>>
>> Avati
>>
>> On Mon, Jan 26, 2009 at 9:48 PM, Filipe Maia <filipe@xxxxxxxxxxxxxx> wrote:
>>> Hi,
>>>
>>> When doing self heal the access and modification times are not
>>> preserved. This can have bad consequences for certain applications
>>> (e.g. make).
>>> The attached patch tries to fix this.
>>> It doesn't set ctime as is impossible to do in a portable way.
>>>
>>> I also noticed that removing the namespace with glusterfsd running
>>> results in glusterfsd getting confused. It has to be restarted for
>>> things to go back to normal.
>>> It would be nice to be able to avoid this.
>>>
>>> Filipe
>>>
>>> _______________________________________________
>>> Gluster-devel mailing list
>>> Gluster-devel@xxxxxxxxxx
>>> http://lists.nongnu.org/mailman/listinfo/gluster-devel
>>>
>>>
>>
>
